AI Act basis
Actual EU AI Act material, translated into product evidence.
Epok is not a legal-text generator. The current demo uses a Loan Review Assistant to show how Regulation (EU) 2024/1689 evidence work becomes product facts, technical artifacts, reviewer decisions, and a coherent AI System record.
The Act creates a harmonised framework for placing AI systems on the EU market and putting them into service. For teams building serious AI products, that means compliance is not only a legal interpretation problem. It is also a data, model, logging, oversight, and documentation problem.
Epok focuses on that second layer: making the evidence trail visible and reusable so human reviewers can make accountable decisions from facts rather than late-stage reconstruction.
Epok supports preparation and review workflows. It does not replace counsel, notified bodies, regulators, or accountable internal approvers.
Regulation map
Evidence, not guesswork
Risk-based classification
The AI Act uses a risk-based framework. Epok starts by turning intended use and deployment context into an AI System classification record.
High-risk evidence work
For high-risk systems, requirements become operational evidence: data governance, technical documentation, logs, oversight, accuracy, robustness, and cybersecurity.
Auditable documentation
Epok generates draft documents from stored evidence and deterministic templates so reviewers can trace statements back to source artifacts.
Timeline
The deadline is not one day. It is a sequence of obligations.
August 2, 2026 is the big product-planning date because the AI Act becomes broadly applicable then. But parts of the framework already apply, and some high-risk product categories have a longer transition. Epok treats this as an implementation timeline, not a one-off document sprint.
Step 1
August 1, 2024
AI Act entered into force
Step 2
February 2, 2025
Prohibited practices and AI literacy obligations started applying
Step 3
August 2, 2025
Governance rules and GPAI obligations started applying
Step 4
August 2, 2026
Broad full application date
Step 5
August 2, 2027
Extended transition for high-risk systems embedded in regulated products
EU fast-track mission
Europe-first AI needs faster proof, not weaker standards.
The EU can compete with the US and China by making trustworthy AI easier to evidence, faster to review, and more repeatable across the single market. Epok exists to make evidence preparation feel native to technical teams instead of bolted on after the model is built.